A tool I reviewed audited my own site and found a bug I'd missed

This captures something that silently kills shipping velocity: the gap between "I noted it as done" and "it measurably works in production."

The og:image thing is sharp because it's not a missing feature - it's a completed task on a checklist. You marked it done, moved on. But your measurement system was the person-who-wrote-the-note, not the automated verifier.

I notice teams that move fast maintain parallel measurement systems: human checklist (fast, forgetful) and automated audit (slower to set up, reliable, finds three more things). The audit doesn't feel like "more work" - it's the difference between "we think we solved it" and "we know we did."

And that gap compounds across 50 features. Your checklist says 50/50 done. The tool says 22/50. Which measurement system determines what you fix next?

The two findings that make this post work are both claimed-versus-actual gaps: your notes said social previews done while Frase found zero tags, and Make logged success while the email never sent. I would make that exact gap a fixed section of every review instead of a side lesson. 'What the tool says happened versus what we could verify' is harder for competitors to copy than opinion, and it gives readers a reusable question instead of another rating. On the traffic question: at position 36 with 19 impressions, I would publish more pages first, but only pages with a distinct verified finding; thin filler will not give you anything worth ranking.

The 'Make logged success while the email silently never sent' line is the one that would haunt me, because a tool reporting success on a partial failure is worse than an honest error. I run automations daily and now treat every green checkmark as unverified until I check the actual downstream artifact, the sent message or the written row, not the tool's own status. On content vs backlinks: at position 36 with single-digit impressions, links mostly lift pages you already rank 5 to 15 for, so more indexed pages usually give Google more surface to find those first. What's the smallest publishing cadence you can actually sustain without burning out?

This is why I paid for a Zarek audit before PH. Found 8 issues, 3 critical. The "expert eye" catches what the builder is blind to after 100 hours in the code.

Best $0 I spent (it was a free community audit). Worth 10x more than any ads.

The part about writing “social previews done” in your notes and then discovering they were completely broken is painfully relatable. 😅

I think that's also why actually using a tool before reviewing it is such a good rule. A feature can look perfect on paper while something completely different breaks once it touches a real workflow.

And honestly, 19 impressions / 0 clicks after only a couple of weeks doesn't sound particularly worrying to me. I'd probably prioritize publishing consistently first, then use GSC to see which pages/queries start getting impressions before going hard on backlinks.

Curious to see how the traffic changes once you've got a bigger sample size.

The "og:image done" → "actually zero tags" gap is pure measurement clarity in action. Your notes were a mental model. The tool was reality.

Most founders never run this kind of audit because it's easier to trust our own annotations. But what you're doing here is outsourcing verification to something that doesn't know how to lie - it just counts what exists.

That 19 impressions / 0 clicks is interesting because it actually tells you something specific: you're getting search visibility but the click-through is catastrophic. Most people would blame "position 36" and build more content. But your stack audit already found a concrete friction point (og:image, and presumably those "success" logging failures).

The measurement system determines what you can see. You built one that surfaces friction instead of just celebrating views.

The “I already marked this as done” part is the most interesting to me. It shows how easily our own notes become assumptions instead of evidence. Automated audits are valuable not just because they find bugs, but because they challenge what we think we've already verified. Curious whether you'll start treating these audits as a recurring pre-publish check.

The strongest part is the willingness to let the product audit the founder’s own assumptions. That makes the “review” model more credible because the process is producing findings, not just content about tools.